minimum change, maximum effect
I like my version better, thank you.
I mean the technicals could probably stored into somewhat manageable functoins, like "if before a certain time internaval, the human player has or has not done this and that -> save as manipulator for following game". but the question is if there is even a brigde for ai's outside of games itself.
In theory it's possible, yeah. The game already has stats for pretty much everything, and in most cases you could probably draw some parallels between the stats and strategy used (stuff like x credits spent on tactical buildings across x planets of the player's empire, slow ship progression = turtler). But then it's not so much a learning AI as an adaptable one, and one that's easy to exploit (win a game by turtling, and know that the AI in the next game will go with a rush to try and get you before you get defenses up).
A real "learning" AI would have to make these decisions during a game.. and that's a whole other ballgame 
